In 1989, to mimic the human BCL2-IGH translocated allele and assess the tumorigenic potential of BCL2 in vivo, Mc Donnell and colleagues developed the first BCL2 transgenic mice bearing a human BCL2-Ig minigene where expression of BCL2 is restricted to B cells (62). The mice developed follicular hyperplasia made of small naive B cells—expressing markers such as B220, IgD, IgM and Igκ—with prolonged survival in vitro. In original studies, after 18 months of age these mice showed high-grade lymphomas although at a low penetrance and interestingly half of those mice harbored c-myc rearrangement. More recent follow-up studies showed that 40% of BCL2-Ig mice develop FL-like tumors expressing GC markers (PNA⁺BCL6⁺) by 17–18 months when chronically immunized with sheep red blood cells (SRBC) over 6 months (87). Strasser and colleagues similarly engineered a transgenic strain Eµ-BCL2, where BCL2 is placed under the control of the 5’ IGH enhancer Eµ (63). This model also showed an expansion of small B cells and plasma cells but did not yield tumor development, however an increased incidence of other B-lymphoid neoplasms was observed (64).

The first experimental model that faithfully reproduced the human disease in term of localization, histology, phenotypic and genotypic features involved expression of Bcl2 in all hematopoietic cells. Egle and colleagues generated the VavP-Bcl2 model where the Bcl2 transgene is controlled by Vav gene regulatory sequences, which confer Bcl2 expression in multiple hematopoietic lineages (65). The mice developed, in 15–25% of cases, isotype-switched, somatically mutated Ig and disseminated lymphomas at 10 months of age. In contrast to Eµ-BCL2 transgenic models of the same age, VavP-Bcl2 mice develop spontaneous expansion of PNA⁺ GC lesions in otherwise ‘healthy’ mice, a premalignant condition which is strongly dependent on CD4 T cell help as in vivo removal of CD4 cells almost abolished GC hyperplasia. In this model, CD4 T cells make a critical input into the exaggerated GC reactions and eventually the onset of FL. A second model developed by the group of Boxer (66) consisted of a IgH-3’E-Bcl2 knock-in mice, where the IgH 3’ enhancer regions was integrated 3’ of Bcl2 locus, thereby mimicking the effects of the long distance IgH 3’ enhancer on Bcl2 expression and limiting Bcl2 expression to mature B cells. In addition to an altered B-cell differentiation and increased B cell numbers in the spleen, lymph node (LN) and BM, these mice recapitulated, between 7- and 14-months, typical histopathological features of GC-experienced FL-like tumors expressing the GC markers PNA and Bcl6 surrounded by FDC networks and CD4⁺ T cells. Despite its main limitations due to the unspecific Bcl2 deregulated expression in all hematopoietic cells, VavP-Bcl2 is currently one of the most popular models to study in vivo the role of secondary genetic alterations that are frequently found in combination with BCL2 in FL and GCB-DLBCL genetically engineered mouse models.

These studies support the notion that despite a survival advantage conferred to B cells by BCL2, the lymphomagenic process requires additional hits (genetic and/or immunological) for transformation.

Although several BCL2-engineered models have provided the initial proof-of-principle that BCL2 ectopic expression leads to FL and high-grade lymphomas, the expression of BCL2 in all B cells (and all T cells for some models) do not represent a true premalignant intermediate stage seen during human lymphomagenesis, where the first t(14;18) event occur in a single B cell in the BM, and is carried on until its ectopic expression in the GC. Instead, pan-B cells BCL2 mouse models generate a polyclonal hyperplasia of naive B cells which is not a known progression step in human FL pathogenesis and as a consequence does not allow studying the early steps of clonal emergence and disease progression. The BCL2^tracer mouse model has been engineered to mimic “sporadic” t(14;18) translocation (8). This transgenic model relies on the introduction of potent RAG recombination sites at the vicinity of an inactivated human BCL2 transgenic minilocus. In this construct, recombination allows to turn on ectopic BCL2 expression in only few B cells, and at the appropriate window of B cell development in the BM pre-B cells. The recombination breakpoints provide unique PCR-based clonotypic markers to study early steps of clonal emergence and expansion in mouse blood/tissues at different time points, allowing a precise analysis of clonal progression kinetics. As expected, the emergence of BCL2⁺ B cells was traced in various tissues at low frequency (1 in 10⁵ to 10⁶) recapitulating the “healthy human t(14;18) carrier” situation. At steady state or after acute immunization with a T-cell dependent antigen, BCL2 alone was not able to drive progression of BCL2-expressing cells into a tumor after 18 months of follow-up, confirming that lymphomagenesis is a stepwise process where premalignant B cells require the accumulation of secondary (epi)genetic alterations to progress into a tumor. However, a chronic immunization protocol with SRBC accelerated genomic instability by allowing BCL2-overexpressing B cells to give rise to memory cells that preferentially underwent iterative rounds of GC entry, allowing multiple rounds of AID-mediated mutagenesis over time to ultimately form premalignant in situ FL structures, the earliest known intermediate preceding human FL. Although this model alone does not form FL lesions, it has been instrumental to propose a revised model of early lymphomagenesis whereby cyclic reactivation of BCL2⁺ memory B cells within new GC reactions would constitute a major pathogenic mechanism facilitating clonal expansion and accumulation of secondary mutations in FL precursors. It is likely that this reactivation in humans operates over decades before clinical manifestation (that we will never reach during the mouse lifespan) and ultimately contribute to the generation of a heterogeneous population of aberrant memory B cell intermediates resembling clonal FL evolving in asymptomatic patients years before diagnosis. Interestingly, this model also led to identify an “immunological 2^nd hit” in FL, departing from the all-genetic, cell-intrinsic concept of lymphomagenesis.

CREBBP and EP300 are highly homologous histone acetyltransferases (HAT) that modify gene expression through H3K27 acetylation at enhancer domains of both histone and non-histone substrates. In 50 to 75% of cases, mutations of CREBBP are missense inactivation of the acetyltransferase catalytic domain, the remaining mutant alleles causing truncation or loss of expression (54, 55, 69, 90). CREBBP and EP300 mutations are usually detected only at one allele and exclusive fashion, which is thought to be explained by the compensatory and redundant role of these enzymes. The precise timing and location where CREBBP or EP300 mutations happens during B cell transformation is still unclear in humans, but there is large agreement from clonal evolution studies and in premalignant FL conditions that CREBBP is an early event (91). Several groups have therefore attempted to investigate the role of hemizygous vs. homozygous Crebbp inactivation on mature B cell differentiation and lymphoma development in vivo using different strategies from early inactivation in all hematopoietic cells, early B-cell specific deletion or GC-specific deletion [Cγ1-Cre (92)], in combination or not with BCL2 transgenic models.

Horton and colleagues studied the consequences of Crebbp deletion in the hematopoietic stem cell compartment thanks to a pIpC-mediated Mx1-Cre recombinase system (93). Those animals developed B cell lymphoproliferative disorders accounting for 29% of all deaths. The B cell lesions localized mostly in the spleen and blood and stained with B220, CD19 and surface IgM indicating their mature B cell origin, however these tumors did not express GC markers. The occurrence of these lymphoproliferative disorders was preceded by the accumulation of lymphoid progenitors characterized by a hyperproliferative state and an altered DNA damage response linked to the loss of p53 activation in the absence of Crebbp. Interestingly, the authors found no enrichment for Bcl6 targets among Crebbp binding sites in their system suggesting different epigenetic changes when Crebbp is deleted in hematopoietic stem cells (HSC). This data suggests that CREBBP mutations acquired early during hematopoiesis may contribute to the emergence of B cell lymphomas although this model might not be relevant for FL or DLBCL progression as the tumors do not exhibit GC features. It also indicates that the timing of mutation acquisition has an important impact on B cell development. In humans, detection of CREBBP mutations in HSCs remains a very rare event (93).

Three others groups investigated the B-cell deletion of Crebbp in combination with BCL2 overexpression to fit with the frequent co-occurrence of the two alterations in human FL and DLBCL (67–69). Jiang et al. recapitulated Crebbp downregulation and Bcl2 overexpression in vivo using a shRNA retroviral infection system of VavP-Bcl2 hematopoietic stem progenitor cells (HSPCs) transplanted into lethally irradiated wild-type recipient mice. They observed an acceleration of lymphoma onset in double-mutant vs. VavP-Bcl2 mice. Lymphoma cells expressed B220, CD19 and IgM and were characterized by somatically mutated Ig locus confirming their GC B cell origin. Importantly, a similar phenotype was observed with a shRNA targeting Ep300. The second model developed by Zhang and colleagues consisted in cohorts of Cγ1-Cre (92) and Cd19-Cre Crebbp^{flox / flox} and Crebbp^{flox / +} animals where inactivation of Crebbp was induced in GC or developing B cells respectively. After a follow-up of 18 months they did not observe any significant difference between Crebbp mutant mice and their littermate controls concluding that loss of Crebbp alone at early or late stages of B cell differentiation was not sufficient to induce lymphomagenesis. The generation of mice crossing Cγ1-Cre Crebbp^{flox / flox} mice with VavP-Bcl2 transgenic mice led to a significant increase in the incidence of B cell malignancies resembling human FL (92% in double mutant versus 61.5% in VavP-Bcl2 controls). Furthermore, the tumors were characterized by a follicular architecture, were largely of GC origin with Bcl6 expression and presence of mutated Ig genes. In an extension of this study and to investigate in parallel how CREBBP and EP300 contribute to normal GC B cell physiology, Meyer et al. established mouse lines carrying single “floxed” HAT genes that are excised only in activated B cells in the GC (Cγ1-Cre) (94). In accordance with previous studies (95), Meyer et al. confirmed that loss of Crebbp in GC B cells led to increased GC formation while Ep300 loss led to an opposite effect with decreased proportion of GC B cells in immunized mice (94). Transcriptomic analyses of purified GC B cells obtained from these two strains revealed that the set of genes modified by Ep300 loss or Crebbp loss was different with expression of DZ transcripts preferentially repressed in Ep300-deleted GC B cells while LZ genes were preferentially decreased in Crebbp-deleted GC B cells providing a mechanistic explanation for the reduced numbers of GC B cells observed in the GC-deficient Ep300 model. The most interesting phenotype comes from the drastic reduction of GC B cells observed when both HAT genes are deleted from the same GC cells. Indeed, the GC response was completely abrogated indicating that the ability of GC B cells to proliferate and differentiate relies on the combined activity of both acetyltransferases likely due to their overlapping and partially redundant functions. Interestingly, CREBBP-mutated lymphoma B cells maintained this dependency toward EP300 enzymatic activity which identify a unique vulnerability that provide exciting opportunities of targeting single mutant CREBBP or EP300 GC-derived lymphomas.

In a third model, Garcia-Ramirez et al. produced strains of Mb1-Cre Crebbp^flox/+ or Crebbp^flox/flox mice where Crebbp was deleted at early pro-B cell stage in the BM. Crebbp deletion at early stages of B cell development led to reduced frequencies of B-cell subsets with reduced numbers of total B220⁺ B cells in BM and spleen. When crossed with the Eµ-BCL2 mouse model, these animals showed higher frequencies of GC B cells in the spleens after SRBC immunization and occasionally develop clonal B cell lymphoma with low penetrance and long latency (13 months). Histology was similar to human FL grade 3 or DLBCL and tumoral cells expressed Pax5 and Bcl6 and displayed SHMs consistent with a GC B cell origin (67).

Overall, these three independent in vivo models confirmed an oncogenic cooperation between CREBBP loss of functions and BCL2 overexpression to promote lymphomagenesis in vivo. Mechanistically, these studies showed that CREBBP, and likely EP300, maintain H3K27 acetylation at certain enhancer that are poised during GC reaction and whose reactivation is required for GC exit and terminal differentiation. This includes genes involved in immune synapse, downstream effectors of BCR and NF-κB or terminal differentiation (Irf4, Nfkb2, Cd40) and antigen presentation, the most notable being MHC class II molecules which are critical for B cell terminal differentiation. In normal GC cells, CREBBP targeted enhancers are direct targets of BCL6 and transiently repressed by BCL6/SMRT/HDAC3 complexes that deacetylate H3K27. Upon selection signals received in the LZ allowing GC exit, these enhancers recover H3K27ac state as CREBBP can directly acetylate BCL6 to inactivate its function by preventing the interaction with co-repressor complexes. By impairing the reactivation of these enhancers leaving unopposed BCL6 oncogenic activity, CREBBP loss of function disrupt the expression of immune synapse genes and their downstream signaling pathways, resulting in accumulation of aberrant GC B cells that fail to properly respond to exit signals from the GC microenvironment thereby promoting lymphoma progression. In human FL, decreased MHC II expression and reduced CD4 and CD8 T cell infiltrations have been described in CREBBP-mutant FL. A similar association between CREBBP inactivation and reduced expression of MHC class II is observed in murine lymphoma models which alters mutant GC cells ability to present antigen to CD4⁺ cells. Interestingly, CREBBP-mutant lymphomas become dependent to HDAC3, the histone deacetylase opposing the effect of CREBBP, that has been identified as a relevant therapeutic target in these tumors (69). Mondello et al. recently showed that HDAC3-selective inhibitors have a dual effect by reversing CREBBP-mutant aberrant epigenetic programming limiting lymphoma growth inhibition while restoring antitumor immunity, notably antigen presenting-genes (96).

KMT2D is a component of the COMPASS complex involved in transcriptional activation through H3K4 monomethylation of gene enhancers in B cells. The majority of KMT2D mutations are nonsense events leading to a truncated protein lacking the enzymatic SET domain involved in H3K4 methylation resulting in a loss of function (90). To study the consequences of Kmt2d inactivation during B cell differentiation, Zhang and colleagues generated a conditional Kmt2d knock-out model relying on a Cre-Lox system with conditional deletion of B cells early during B cell development (Cd19-Cre Kmt2d^flox/flox ) or after GC initiation (Cγ1-Cre Kmt2d ^flox/flox). Early deletion of Kmt2d lead to a higher number and enlarged GC formation while this effect was not seen with late GC deletion and in vitro, Kmt2d-deficient cells displayed a proliferative advantage compared to wild-type cells. However, Kmt2d deficiency in B cells alone upon early or late inactivation was not sufficient to induce FL or DLBCL in vivo. Kmt2d protein was mainly found on putative GC enhancers and global H3K4 methylation levels were diminished in Kmt2d mutant mice. Moreover, although GC-specific deletion was insufficient to initiate malignant transformation, Kmt2d^ko -VavP-Bcl2 double-mutant mice developed B-cell lymphoproliferative disorders with an incidence of 78% (44% for VavP-Bcl2 alone), with tumors expressing BCL6 and PAX5 consistent with a GC origin and recapitulating a spectrum of histopathological features ranging from early FL to DLBCL (40% early FL, 31% FL and 27% DLBCL) (70).

Using a different experimental system, Ortega-Molina and colleagues also explored cooperation between bcl2 and Kmt2d deletion in B cell lymphomagenesis. Using a retroviral infection system of shRNAs transduction to silence Kmt2d in HSPCs from VavP-Bcl2 donor mice following transplantation into irradiated wild-type mice, an acceleration of the lymphomagenesis process as well as an increase in the incidence of FL-like tumors (from 30 to 60%) was observed in double-mutant mice compared to shRNA control vectors, validating the tumor suppressive role of Kmt2d in B lymphocytes. At preclinical stages of the disease, Ortega Molina et al. showed that after immunization, the number of GC B cells was increased when Kmt2d was suppressed. Moreover, GCs persisted for a longer period than in control mice. Kmt2d loss was associated with a decrease in IgG1 production suggesting a dysfunction of the class switch recombination processes (71). Interestingly, the generation of Kmt2d^flox/flox Cd19-Cre crossed with a strain overexpressing AID led to the development of aggressive lymphomas resembling DLBCLs and confirms that, independently from BCL2 expression, genetic instability linked to AID overexpression cooperates with Kmt2d loss to promote lymphomagenesis.

Integrative genomic analyses from human samples carrying KMT2D mutations and Kmt2d-mouse FLs showed that genes differentially expressed in Kmt2d-mutated lymphomas were mostly repressed and affected a set of genes involved in terminal differentiation programs and GC exit, such as CD40 and BCR signaling, regulation of apoptosis, control of cell migration and proliferation. KMT2D mutations result in persistent demethylation of enhancers and failure of the respective genes to respond to signals, notably CD40 signaling from T_FH. Moreover, absence of Kmt2d affects negatively the expression of major B-cell tumor suppressors such as Tnfaip3, Socs3, Tnfrsf14, Asxl1 or Arid1a. In conclusion, lack of Kmt2d leads to an aberrant repression of key genes normally required for GC exit, favoring an abnormal GC B cell outgrowth and failure to differentiate leading eventually to lymphoma development. Of note, these studies showed once more that the stage and/or the timing of a given (epi)genetic alteration has a strong influence on the transcriptional changes occurring in the GC. The developmental stage at which Kmt2d mutations are introduced in human precursor tumor cells is still unknown but it has been hypothesized that epigenetic reprogramming may require multiple rounds of cell divisions to allow the replacement of modified histones by non-modified histones explaining why Kmt2d inactivation after GC initiation may have a more modest phenotype than early inactivation.

EZH2 is a H3K27 methyltransferase part of the Polycomb Repressive complex-2 (PRC2). Heterozygous gain of function mutations, preferentially affecting the EZH2 SET domain at the Y641 residue and making EZH2 more efficient at H3K27 trimethylation (97), are found in up to 30% of GCB-DLBCL and FL and de facto enriched in EZB/C3 DLBCL subtype (18, 89, 98). Several groups have investigated the functional role of wild-type and mutant EZH2 during GC reaction and B cell lymphomagenesis (73, 75, 99, 100). Using a Cγ1-Cre Ezh2^flox/flox strain, Béguelin et al. observed a marked reduction in GC B cells after immunization. They reproduced this phenotype in immunized wild-type mice treated with an EZH2 inhibitor targeting wild-type and mutant EZH2, establishing that EZH2 is required for GC formation (73). Similar observations were made by Caganova et al. underlining that under normal conditions, EZH2 enables GC formation at least in part by suppressing cell-cycle checkpoint genes like CDKN1A, impairing DNA damage responses to support centroblast proliferation and silencing essential plasma cell differentiation genes, particularly Blimp1 and Irf4 (99, 100).

To understand how EZH2^Y641 hotspot mutation perturbs GC development and drives lymphomagenesis, Béguelin and colleagues developed two mouse strains conditionally expressing the mutant Ezh2^Y641N or Ezh2^Y641F in GC B cells upon Cγ1-Cre recombinase (73, 75). Upon immunization, both models caused oversized GCs and displayed increased abundance of H3K27me3 mark at critical GC B cell bivalent promoter leading to permanent silencing of EZH2 target genes. The GC phenotype appears to be mediated through cooperative and mutually interdependent actions of EZH2 together with the transcriptional repressor BCL6 and BCOR repressive complex. These Ezh2 mutant mice did not develop lymphomas. However, early activation (CD19-Cre) of mutant Ezh2 in an independent study led to aggressive DLBCL in about 12 months (72). VavP-Bcl2 HSPCs were transduced with retroviruses expressing Ezh2^Y641F , Ezh2 ^WT or an empty vector and transplanted into lethally irradiated recipient mice subjected to SRBC immunization every 4 weeks to induce GC formation. Ezh2^Y641F Bcl2 ⁺ chimeric mice led to early lymphoma development in 70% of mice at 111 days (vs 20% in mice overexpressing Ezh2 ^WT Bcl2 ⁺ and none in the Bcl2 control at that stage), characterized by enlarged spleen and liver and resembling morphologically to DLBCL with centroblastic morphology (73, 74). These data demonstrate that EZH2 gain of function mutations accelerate GC B cell lymphomagenesis in cooperation with Bcl2 overexpression, recapitulating features of GCB-DLBCL. Similar results were obtained when transgenic and knock-in Ezh2 strains engineered to express heterozygous mutant Ezh2 in GC B cells were crossed with VavP-Bcl2. Importantly, homozygous expression of mutant Ezh2 phenocopies the Ezh2 knock-out phenotype further attesting the requirement for the maintenance of the wild-type allele for Ezh2 mutant enzymatic activity.

Recent studies questioned whether EZH2 mutation has additional and qualitatively distinct function in lymphomagenesis beyond simply being a more potent version of the wild-type enzyme. Along these lines, Ennishi et al. identified a strong enrichment of EZH2 mutations in human DLBCL cases with loss of MHC-I and MHC-II expression linked to a reduced number of tumor-infiltrating lymphocytes and less T cell cytolytic activity (74). To investigate in detail the consequences of EZH2 mutations on MHC expression and immune microenvironment, the authors relied on two different mouse models. First, they used the Ezh2-mutant model developed by Béguelin et al. where VavP-Bcl2 hematopoietic progenitors are infected with Ezh2-mutant containing retrovirus before re-injection into lethally irradiated recipient mice. Second, they used Cγ1-Cre Ezh2^Y641N or Ezh2^Y641F mice crossed with VavP-Bcl2 strain which develops DLBCL-like tumors. In both experimental systems, MHC-I and MHC-II expression were significantly reduced in mutant Ezh2 mice compared to wild-type mice, with reduced infiltration of CD3, CD4 and CD8 T cells in the tumor microenvironment, establishing Ezh2 gain-of-function mutation as a driver of MHC downregulation in GC lymphomagenesis, and eventually favoring immune escape. Importantly they also showed that EZH2 epigenetic switch-off of MHC molecules, driven by transcriptional repression of MHC-I/II transactivators, could be reversed with EZH2 inhibitors. The ability to restore MHC expression provides an interesting proof of concept in combining epigenetic reprogramming small molecules with immunotherapeutic approaches.

More recently, Melnick and colleagues further strengthened the concept that Ezh2 mutation initiate GC derived lymphomagenesis by escaping immune effector recognition and inducing a remodeling of the GC immunological niche (101). Using competitive BM chimera to track the cellular dynamics of Ezh2^Y641F B cells during GC reaction, they found that Ezh2 mutants manifested a competitive growth advantage in the GC. This competitive advantage led to GC hyperplasia characterized by an increase of a cycling LZ cell population, without maturation blockade, associated with the expansion of FDC network. Using droplet based single cell transcriptomics and advanced histone mass spectrometry technologies, they assessed how Ezh2 mutation affected GC B cell cellular and histone methylation dynamics in an unbiased manner. They found that the epigenetic reprogramming imprinted by Ezh2 mutation through the reinforcement of the repressive program induced by H3K27me3 accumulation led to the abolition of LZ B cells’ dependence toward T_FH emanating signals. Indeed, DZ re-entry of Ezh2-mutated centrocytes was clearly diminished and these GC B cell escaped T_FH-mediated clonal selection. The ability of Ezh2-mutant GC B cells to proliferate within the LZ was linked to FDC interaction as this ability was impaired when FDC function was abolished through injection of soluble lymphotoxin receptor β. Remarkably, this study revealed how activating mutation of EZH2 induces a premalignant FL-like niche allowing B cells to persist as slowly proliferative centrocytes without T_FH help and in a FDC dependent manner.

The above studies provide important insights on an emerging paradigm where one of the most critical function of epigenetic modifier mutations in promoting GC B cell transformation goes beyond the sole reprogramming of the GC epigenome, but instead arise from failure of GC exit signals to restore expression of genes that normally regulate immune signaling pathways and antigen presentation (47). Of therapeutic interest for the targeting of the FL common precursor cells, this remodeling of the immune synapse at least with EZH2 mutations tends to occur early during lymphomagenesis.

Linker histones (H1) are additional chromatin modifying genes involved in the organization and stabilization of the nucleosome structure, supporting the folding of chromatin into higher-order conformation, and regulating its epigenetic state through the recruitment of histone modifiers. Heterozygous H1 mutations, found in up to 44% of FL and 27% of GCB-DLBCL, are mostly missense events affecting the globular C-terminal domain which led to the loss of protein function with impaired chromatin binding (53). Among them, H1C and H1E are the most common affected H1 isoforms observed in B cell lymphomas (102, 103).

To investigate the functionality of these isoforms in lymphomagenesis, Yusufova and colleagues used the previously described H1c^−/− H1e^−/− mice model (104). At early time points after chronic SRBC immunization, those animals manifested lymphoproliferative disease with invasiveness of B220⁺ B cells in extranodal tissues such as liver and lungs. Further analysis using competitive BM chimera revealed a competitive advantage for H1c/H1e-deficient B cells characterized by a specific increase of cycling LZ B cell population expressing Gl7, Fas and cd86 markers whereas no effect was found in other mature and immature B cell populations. Notably, H1 deficiency enables a chromatin decompaction in GC B cells with an enrichment in stem cell genes that become desilenced in H1-deficient GC cells. Given the frequent co-occurrence of H1C and H1E mutated genes with BCL2 overexpression in lymphomas, they crossed H1c^−/− H1e^−/− mice with VavP-Bcl2 mice. Loss of H1 isoforms caused a more extensive disruption of lymph node architecture with diffuse infiltration of immunoblastic cells, along with an extensive invasion of B220⁺ B cells and CD3⁺ T cells in extranodal tissues establishing H1 proteins as bona fide tumor suppressors. RNA sequencing analysis of lymphoma-like tumors in mouse and humans revealed a significant enrichment for stem cell signature and serial transplantations confirms that loss of H1 conferred lymphoma cells with enhanced self-renewal potential. These findings enlighten the contribution of H1 linker deletion in driving malignant transformation where epigenetic marks changes favor a relaxed state chromatin in GC B cells, increasing B cell fitness advantage by allowing self-renewal proprieties and may expose DNA to further AID-mediated additional hits.

Besides chromatin modifiers genes, a number of studies have demonstrated methylation and disruption of cytosine methylation [5-methylcytosine (5mC)] patterning as another factor linked to the biology of B cell lymphoid malignancies (105). The methyl-cytosine dioxygenase TET2 (ten-eleven translocation 2) missense or truncated mutation is present in 6–12% of GCB-DLBCL (14). TET2 mutations are known to occur early in human HSCs and can be found in individuals with clonal haematopoiesis. Whether early TET2 mutations has a driving role in DLBCL was explored in multiple models. 5mC is well established as an epigenetic mark associated with transcriptional silencing, notably of tumor suppressor genes. TET2 is involved in active DNA demethylation, catalysing the oxidation of 5mC to 5-hydroxymethylcytosine (5hmC). Recently, it has been appreciated that 5hmC also functions as an epigenetic mark, and when linked to gene enhancers, is associated with activation of nearby genes.

Conditional deletion of Tet2 specifically in the B cell compartment with CD19-Cre Tet2 ^fl/fl mice showed B cell transformation mimicking chronic lymphocytic leukemia (78). Programmed deletion of Tet2 in hematopoietic cells (Vav-Cre) or B cells (CD19-Cre) in immunized animals disrupt the ability of GC B cells to undergo CSR and terminal differentiation. Furthermore, conditional deletion of Tet2 at the GC stage results in a preneoplastic GC hyperplasia, blockage of GC exit and PC differentiation evolving in DLBCL-like tumors, confirming its role as a bona fide B cell tumor suppressor (79). Mechanistically, this phenotype is due to the focal loss of 5hmC at enhancers linked to B cell differentiation. Indeed, Tet2^−/− GC B cells feature disruption of many enhancers linked to GC exit signaling pathways, antigen presentation, and terminal differentiation genes. This mechanism is conceptually similar to the functions of the histone modifiers in DLBCL which fails to restore the immune synapse. Interestingly, TET2 and CREBBP mutations are mutually exclusive in DLBCL (106), thus a combined mouse model could be engineered to find a potential therapeutic vulnerability in DLBCL.

The methyl-cytosine dioxygenase Tet1 (ten-eleven translocation 1) is also an important regulator of 5-hydroxymethylcytoine and interestingly transcriptionally silenced in FL. Cimmino and colleagues engineered Tet1-deficient mice where B cell lymphoma development was promoted resulting in a diminished survival compared to wild type mice (77). Tet1-deficient mice exhibited lymphadenopathy and hepato-splenomegaly. Splenic tumors were characterized by a massive infiltration of proliferating lymphocytes disrupting the normal architecture and expressing the GC markers Bcl6 and Irf4 but not the PC marker Cd138. When combined with Bcl2 overexpression, Tet1-deficient B cell lymphomagenesis was accelerated up to 10 weeks post-transplantation. Altogether, deletion of Tet1 and Tet2 in mice induces phenotypically predominant DLBCL tumors supporting a suppressor role in mature B cells.

TNFRSF14, the gene encoding the HVEM receptor located on 1p chromosome is one of the most frequent cell surface protein, deleted or mutated in >40% of FL cases and enriched in GCB-DLBCL (14, 18, 19). HVEM is expressed at the surface of B cells as well as other cell types and has multiple ligands including LIGHT or BTLA. Besides its role as a signaling receptor, HVEM can act as ligand and transmit signals into BTLA-expressing cells notably T_FH (107). How loss of HVEM contributes to lymphomagenesis has been the focus of two independent and complementary in vivo studies (80, 108). The first study took advantage of the well characterized VavP-Bcl2 model that recapitulates key aspects of the genetics and pathology of human FLs to generate BM chimeras where knockdown of Hvem was mediated by transduction of shHvem into VavP-Bcl2 HSCs followed by reconstitution of irradiated wild-type hosts. Knockdown of Hvem in all hematopoietic system caused a significant acceleration and increased penetrance of lymphoma development compared to VavP-Bcl2 controls with 90% of animals carrying tumors at 100 days. Despite the absence of B-cell specificity of the shRNA transduction strategy, only B cells were enriched with the short-hairpin construct indicating the Hvem knockdown in T cells was unlikely to participate to the lymphomagenic effect. Mechanistically, besides cell-autonomous activation of B cells, Hvem loss and the consequent loss of interaction with Btla triggers the amplification of T_FH producing high amounts of TNF and Lymphotoxin, the two non-redundant factors involved in lymphoid stromal cell differentiation and maintenance, and favors lymphoid stromal cell activation including FDC and FRC (Follicular Reticular Cells). This seminal study offers the first demonstration of a functional impact of a B-cell specific genetic alteration on the polarization of a FL-supportive microenvironment (80). Of therapeutic interest, immunotherapeutic delivery of a soluble HVEM receptor, via modified CAR-T cells, inhibited the growth of lymphoma by restoring the BTLA-HVEM interaction highlighting the essential role of the tumor/microenvironment dialog in lymphomagenesis. In an independent study, Mintz et al. reported that in models of Bcl2 overexpression in B cells, Btla deficiency in T cells led to a similar GC B cell outgrowth and accelerated lymphomagenesis than Hvem deficiency in B cells proposing an alternative mechanism by which the Btla-Hvem axis functions as an cell-extrinsic suppressor in lymphomagenesis (108). Using a chimeric mouse system, they identified that during a normal immune response, Hvem restrains B cell proliferation, differentiation and selection by reducing the delivery of signal in trans through the Btla-Hvem axis on T_FH cells. Hvem mutation in B cells would lead to a loss of negative signaling in T_FH cells and allows Hvem-mutant B cells to receive exaggerated helper signals that promote proliferation and accrual of AID-mediated mutations. Collectively, those data provide important evidence for a cell-extrinsic tumor suppressor role of Hvem. The ways in which increased signaling via CD40 and other T cell–derived helper factors cooperates with Bcl2-overexpression in lymphoma development remain to be fully elucidated.

Cathepsin S (CTSS) is part of a family of cysteine proteases whose role is essential in the regulation of normal immune response through its activity on antigen processing, B cell expansion and communication with CD4⁺ T cells. By cleaving the CD74 chaperone protein bound on MHC class II molecules, CTSS enzymatic activity results in a smaller peptide CLIP that will be displaced and allow variable antigenic peptides to bond to MHCII and present at the cell surface. Recurrent hotspot mutations and gene amplifications of CTSS have been recently described in 6 and 13% of FL patients respectively (81, 109) and mechanistically, CTSS^Y132D hotspot mutation promotes activation of the protein and increases its protease activity. To assess how the most common CTSS^Y132D mutations or CTSS overexpression contribute to accelerate lymphoma development, Dheilly et al. generated a chimera mouse model of FL using the VavP-Bcl2 mice as HSPC donor cells and expressing either mutated human CTSS or overexpressing human CTSS. Both models revealed an oncogenic role of CTSS over-activation with higher penetrance and decreased latency as compared to VavP-Bcl2 tumors alone. Tumors with CTSS alterations were characterized by a remodeled tumor-prone microenvironment with an increased infiltration of CD4⁺ T cells while limiting CD8⁺ T cells recruitment. Depletion of CD4⁺ T cells in VavP-Bcl2/CTSS chimera models confirm that CTSS is essential to support the communication and co-stimulatory signals between tumor B cells and CD4⁺ cells in the GC context. Conversely, loss of CTSS activity in aggressive mouse lymphoma xenograft restrain lymphoma growth by recruiting and enhancing CD8⁺ T cells cytotoxic activity while impairing communication with CD4⁺ T_FH cells. These data show for the first time that by altering the processing of antigenic peptides, CTSS mutations or overexpression remodel the immune microenvironment to promote lymphoma growth and implies that targeting a regulator of antigen presentation such as CTSS could modulate the spectrum of processed antigens, promote activation of cytotoxic T cells, enhance tumor immunogenicity and improve response to anti-PD1 immunotherapies.

Another pathway frequently mutated in GC-derived lymphomas is the GC homing pathway involving S1PR2 and GNA13. The guanine nucleotide binding protein GNA13 (encoding Gα13), is a signaling mediator downstream of transmembrane G-protein-coupled receptors sphingosine-1-phosphate receptor-2 (S1PR2), that confines B cells in the GC and promotes growth regulation by suppressing both Akt and cell migration (110). In about 20% of GCB-DLBCL cases (38% in the EZB subtype), deleterious mutations affect one of the members of the Gα13 homing pathways, namely GNA13, S1PR2 and P2RY8 (another S1P receptor expressed on GC B cells). To model in vivo the impact of GNA13 loss during lymphomagenesis, Muppidi and colleagues utilized a mixed BM chimeric approach, deleting Gα13 in all B-lineage using Mb1^Cre;Gα13^f/f mice. Deficiency of Gα13 favors the formation of enlarged mesenteric LN with GC B cells expansion associated with a marked disruption of the GC architecture and a loss of GC confinement due to the inability to suppress migration in response to S1P. Occasional transformation in B cell lymphomas displaying a GC-like phenotype (Gl7⁺CD138^-Bcl6⁺Irf4⁺IgD^-) were observed at 1 year of age (82). One of the critical observations made in Gα13 deficient cells is the loss of confinement which allows egress outside the GC, dissemination and seeding of these tumoral cells at distant sites such as blood and BM. As deficiency in S1PR2 did not phenocopy Gα13 deficiency, the authors searched for additional Gα13 G protein coupled receptor that may be involved in GC B cell regulation and discovered that P2RY8, an orphan receptor also represses GC B cell growth promoting confinement via Gα13 and is mutated in GCB-DLBCL. Cooperation between Bcl2 overexpression and Gna13 loss showed an exacerbated phenotype in double mutant mice leading to greater accumulation of GC B cells in spleen, wider dispersal throughout the follicles and more dissemination in blood and BM suggesting a combinatorial effect of Bcl2 in promoting abnormal B cell survival outside the GC niche. These findings shed new lights on an important mechanism by which disruption of Gα13 signaling exerts dual actions in promoting growth and favoring dissemination of GC B cell in GC-derived lymphomagenesis and offer a biological explanation for factors leading to systematic dissemination of tumoral GC B cells in multiple organs including the BM.

The RRAGC gene encodes for Ras-related GTP-binding protein (RAGC), an essential activator of mTORC1 downstream of the sensing of cellular nutrients. Recurrent point mutations of RRAGC are found in 18% of FLs and are rare in other B cell lymphomas (112, 113). More than half of these events are associated with mutations of certain components of the v-ATPase complex (ATP6V1B2 and ATP6AP1). To characterize the impact of recurrent RRAGC mutations on B cell function and lymphomagenesis, Ortega-Molina et al. used CRISPR/Cas9 genome editing techniques to engineer two independent murine Rragc knock-in mouse strains reproducing hotspot mutations recurrently found in human FL (S75C and T90N mutations) (83). Heterozygous RagC^S74C/+ and RagC^T89N/+ heterozygous mice showed no obvious phenotype. Lymphoid and myeloid cell populations frequencies were similar in mutant versus wild-type mice in the spleen and BM however RagC knock-in mutations conferred partial insensitivity to nutrient withdrawal. When bred to a VavP-Bcl2 mice, RagC-Bcl2⁺ double-mutant mice showed exacerbated B cell responses in response to immunization characterized by enlarged GCs, increased plasma cell production without impairment of high-affinity B cell selection and eventually acceleration of lymphoma development. This was observed both in the progeny of VavP-Bcl2-RagC^mut mice but also using a BM chimeric system where VavP-Bcl2-RagC^mut fetal liver cells were transplanted into lethally irradiated wild-type recipients. Histological analysis of these tumors revealed a follicular growth pattern of Bcl6⁺ B cells with no difference according to genotype. Bulk RNA sequencing analysis of B220⁺ cells obtained from RagC^mut and RagC^WT Bcl2⁺ FL-like tumors revealed that the mTORC1 signaling signature was enriched in RagC^mut FL. Furthermore, when transcriptional profiles from murine and human FL with or without RagC mutations were compared, upregulated genes in murine RagC^mut FL were enriched in RRAGC mutated human FL. Interestingly, the mTOR inhibitor rapamycin, given orally to RagC^mut and RagC^WT Bcl2⁺ mice during long term, inhibited proliferation selectively in RagC^mut FL. Overall, this data shows that RRAGC mutations result in mTORC1 activation regardless of the intracellular amino acid concentration which confers a selective advantage to GC B cells. In addition, the presence of these mutations makes GC B cells less dependent to T_FH signals, as blockade of T cell help through anti-CD40L after GC induction makes RagCmut B cells intrinsically resistant to apoptosis despite T_FH suppression. Finally, when these mutations are associated with Bcl2 overexpression, the lymphomagenesis process is accelerated in a cell-intrinsic manner where mutant GC B cells with increased fitness could continue to undergo cycles of selection and proliferation favoring further acquisition of additional hits (83).

SESTRIN1 has been identified as a target gene of the recurrent 6q deletion in FLs and it is also epigenetically inactivated by EZH2 gain-of-function (84). SESTRIN1 is a mTORC1 regulator that inhibits cell growth when TP53 is activated in response to DNA damage. Inactivation of SESTRIN1, by del6q or by the EZH2 mutation leads to mTORC1 activation. Therefore, SESTRIN1 loss represents an alternative to RRAGC mutations that maintain mTORC1 activity under nutrient starvation. To recapitulate SESTRIN1 deficiency in vivo, in combination with BCL2 overexpression, Oricchio and colleagues retrovirally transduced VavP-Bcl2 HSPCs with shRNAs targeting Sestrin1. Sestrin1 deficiency led to an acceleration of the Bcl2-mediated lymphoma, establishing his tumor suppressive role in B cells. Morphologically, BCL2/Sestrin1-deficient tumors resembled FL with a follicular architecture, they displayed markers of a GC phenotype (PNA and BCL6), showed evidence of somatic hypermutation, an increased tumor proliferation based on Ki67 staining. Remarkably, pharmacological inhibition of EZH2 promotes SESTRIN1 re-expression and restores its tumor suppressive activity, suggesting the possibility to epigenetically control mTORC1 activity in lymphoma. Interestingly, EZH2/RRAGC gain of function mutations and SESTRIN1 loss are mutually exclusive suggesting that these alterations are involved in mTORC1 activity maintenance allowing tumor cells to escape proliferation inhibition (84) also becoming less dependent to T_FH signals as shown in the context of RRAGC and EZH2 mutations (83, 101). In summary, mTORC1 pathway activation, through various mutually exclusive molecular alterations, is clearly involved in B cell lymphomagenesis process and may be associated to a specific sensitivity to mTOR inhibitors.